The combination of both not only promises a more compact and maintainable code for graphics, vision, robotics and other scientific and engineering applications, but also automatically exploits parallelism on GPU or combined computing unit (APU) through OpenCL [8] or CUDA [9].